update axmodel and demo

.gitattributes

@@ -37,4 +37,4 @@ main_axcl_aarch64 filter=lfs diff=lfs merge=lfs -text
 main_axcl_x86 filter=lfs diff=lfs merge=lfs -text
 main_prefill filter=lfs diff=lfs merge=lfs -text
 qwen2.5-0.5b-gptq-int4-ax650/qwen2_post.axmodel filter=lfs diff=lfs merge=lfs -text
-*.axmodel filter=lfs diff=lfs merge=lfs -text
+*.axmodel filter=lfs diff=lfs merge=lfs -textmain_ax650 filter=lfs diff=lfs merge=lfs -text

README.md

@@ -17,7 +17,7 @@ This version of Qwen2.5-0.5B-Instruct-GPTQ-Int4 has been converted to run on the
 
 This model has been optimized with the following LoRA: 
 
-Compatible with Pulsar2 version: 3.4(Not released yet)
+Compatible with Pulsar2 version: 4.2(Not released yet)
 
 ## Convert tools links:
 

config.json

@@ -0,0 +1,51 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 896,
+  "initializer_range": 0.02,
+  "intermediate_size": 4864,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 14,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 2,
+  "quantization_config": {
+    "batch_size": 1,
+    "bits": 4,
+    "block_name_to_quantize": null,
+    "cache_block_outputs": true,
+    "damp_percent": 0.01,
+    "dataset": null,
+    "desc_act": false,
+    "exllama_config": {
+      "version": 1
+    },
+    "group_size": 128,
+    "max_input_length": null,
+    "model_seqlen": null,
+    "module_name_preceding_first_block": null,
+    "modules_in_block_to_quantize": null,
+    "pad_token_id": null,
+    "quant_method": "gptq",
+    "sym": true,
+    "tokenizer": null,
+    "true_sequential": true,
+    "use_cuda_fp16": false,
+    "use_exllama": true
+  },
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": 32768,
+  "tie_word_embeddings": true,
+  "torch_dtype": "float16",
+  "transformers_version": "4.39.3",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

main_ax650

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f19ddeb193769b16aa8c5d9bba887558aa0a4ed10eb50a19d9bc117f1ba527e5
+size 985352

main_axcl_aarch64

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bb111fc00c54abb6142a8f44df087bf104c8150a1cefa6be55c6b174b932c4ec
-size 999008
+oid sha256:1f9f1a1ca329b47f70840e8b6d104ce8248a82326aa2402bccb31144590a8fb2
+size 1725008

main_axcl_x86

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6cba0be8df523e351789fcfa745772423096f52a3b0a760f8d8c9f5b8bb2ec82
-size 1022384
+oid sha256:928d36be31c15d081a7d346464f41458e9624d8b68d5f7dfb3d3189686ce2754
+size 8421624

qwen2.5-0.5b-gptq-int4-ax650/model.embed_tokens.weight.bfloat16.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4e96b0df6d274768cbb7e72404011853d23349999b658dc2f4dfb3c431ea223f
+oid sha256:4a00e7baebfd01831d3104348802fb85cc075365a86b6b0007515fa1b9225e7d
 size 272269312

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l0_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef04ff155507e088023f550c844cf18aa3390246015e2a4585b5d7252d884cce
-size 9061930
+oid sha256:c335becab00a3e60c0c11c1fae0d97e5945b92bf5555b047fbf3d3d45f5fa06e
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l10_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ee27f1c903adc6dc97223f4a22bcd56f52f688c53412949706a36cf1ff03cb7d
-size 9061930
+oid sha256:dc334ccd85edc693eae0793d38aaca9bfeb411099b88f24af87396b9e2d3dd7f
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l11_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad4785e11c1748d2425d1a449034059b3a48910cc258860142d94b97a5de7157
-size 9061930
+oid sha256:d7e36187af29691c533346b0e20ef7575f89ffa48783b323827f58e4c0b4963a
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l12_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5da07b1f21d5974f7b80b77204b9981c5ea0d999ddb2874ad797aea9c1643d84
-size 9061930
+oid sha256:e7a755f6a4121ee94f1903620237ae195a8628702fdb50de26162f536c399d07
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l13_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1750a75cabcdded329b8141a20066741255e0768c8a1e8cd084c0ab10cc5eedf
-size 9061930
+oid sha256:a8216164c2518651e754e26dac698e4a85d677742d01afe367844931f877083c
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l14_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6214f2d1839ef7b8f0d66433e3b5df4e8faef76d31c826a0346959efbbab2e1
-size 9061930
+oid sha256:ae8e1e935534654142c588b9ee5fef648eb83559d83f1391d4af148e9fcf0505
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l15_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:064ad4d2420decc77323bd612e47a249fb884c861facb28e3bba1615c4106d8b
-size 9061930
+oid sha256:9975bb26552f7983515417fe8d2338d33a4d2d960d393e9e6833192992fc8bea
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l16_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1c6be483387fa391dd83ab4747106c12e8d4518f5fc125a896d70fc6a68500c3
-size 9061930
+oid sha256:7713922f7c7c80cadb9aef4e5d8edac42e5f667ad9e33fdf37390ce97da3c5f5
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l17_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f537ef54089b5dc77cde017009cd2c618d0551987a6b1e2f2d9d73686c685a4f
-size 9061930
+oid sha256:bec5406b9c1f2b4f7909e1b6614e29a9653ae754f29518b7abf072a24620b6fc
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l18_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:18af0848c1c8ca5fd9ac797fa449519721cfda08f5f0a5bb0a413a0638d53446
-size 9061930
+oid sha256:6ae94ef72d6b4b7a60eb7c40aab314e8d587f3e88455e39a2b700e7dd2c03205
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l19_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0563eeda6b066e11a999012e2ae1a8e5c49689afdcc031d20a99c4e94a2804b5
-size 9061930
+oid sha256:9bb02c80c9ddc8bd251e064b44ca17a8ce516e845038aca79ed6c9e24727114f
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l1_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ddfe42379d84ab9c803bf03da48c316abf907b7ba42dfb17d1a25ce9a6d1e3df
-size 9061930
+oid sha256:f6a080891672734fc64036bd629477719e2569a6afda3e4e316b61ff86364c83
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l20_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:866d1796031b1b370b2c77d4fe15e33248edad955469d8f05bba9bea59208cac
-size 9061930
+oid sha256:69f3be8f72b6eef91ac0cf91744afe085e148e4fa1f6c4d7b0d718ffb9f469fe
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l21_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5e6e7ddd4533fdd1bcee4be9f492c4572cfafa1bd71f57205178c8567889ccaa
-size 9061930
+oid sha256:6c6cf9c7079bc1758e9f7591b95899c8bbeec6fb2e33f5f512913f4f9a8441ad
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l22_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2fd97d6ea9ccce9fe7c11f11469762a01c4a80c066822d51efd4982fa495e00b
-size 9061930
+oid sha256:8019c3f30ef46f4a04c59e35b3522f1f6692c448de8785325280c6cc5ebdccea
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l23_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:10390e63263955edd522a544dbee058262979c2995d831e6e75ac2c60f715096
-size 9061930
+oid sha256:a2985914bbb2b87fa0fb763ed7a1216af94ec3e8db141ee4e58cd30e802d9a28
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l2_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:72bda9ed6a4b256be005f607347345a086a5b528fc724426c81d60175281e8a7
-size 9061930
+oid sha256:37f1e0a303566f121346af3f88209f2bdffff62b77d1276a8c6b607f92661a2c
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l3_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f84762b41969a9abcba349b057bd27cce883898eba9db22442cd904a5a16080
-size 9061930
+oid sha256:1396fe073b577feb67af43b0d40a7dfbeeea646d29d73d5ea2250137b4041c61
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l4_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:367a6976dac3fbeac8ade0c948608cf97c9362964e78ab3f2381bb0b02a2797a
-size 9061930
+oid sha256:e27b2ad022d2c688aba65ba1582cdfc1225f3021af53d6f9eec433d6128733c5
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l5_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f6b37947e64d3aebf8b70b02ae187a03c8ca776adff199a20f0716f20200bea
-size 9061930
+oid sha256:45c93a96f4ebcff7a7b01d229946645346ba40b1d340f4d66d97bf8dee7c5827
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l6_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:21a801a429a82ecd50fc00bbd68b2df92a505691e7fdbe01fd747f55571d6e01
-size 9061930
+oid sha256:7d580d6a8cdfdc270a1a76276a1c4922a638b01aef7d841c7d291f425873cc35
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l7_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7fedacd5a7ed5ef9d958876885ee38f790891c2b2bdb0937c8a83b7d3fa8e0a
-size 9061930
+oid sha256:80620239eee1edb2980900d8b97c6ed9b2c27e9e0b8e97368df3ab17b8e785a9
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l8_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c55113e53dc5bcf1dea75d6289dc6d42942cfa6fd2fc836fee16b04cc57e56a
-size 9061930
+oid sha256:6e910d13e4e88dd3ad0c7461af7d6151e877bb08fd730957225b2eb0a9743172
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l9_together.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b4fce55f0e43b31ee933e46899488d194296211f0df55390e89bc2dc232c367c
-size 9061930
+oid sha256:00b8126bed769a8f7cfc7e5b4bf4239e111e03168aa15e10bf83c959e4e31cfa
+size 12286102

qwen2.5-0.5b-gptq-int4-ax650/qwen2_post.axmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0e0533f02e303b9d26ab40a1648917d22d9ae7ac3573a24e4b850fb6d0f71e85
-size 147955255
+oid sha256:af27ff530b80e60e1472c89017bb9c981f76960f6b01ac18aa32b3c83f7511fb
+size 147954839

qwen2.5_tokenizer_uid.py

@@ -0,0 +1,199 @@
+from transformers import AutoTokenizer, PreTrainedTokenizerFast
+from http.server import HTTPServer, BaseHTTPRequestHandler
+import json
+import argparse
+import uuid
+
+# 全局字典：存储 uid 到 Tokenizer_Http 实例的映射
+tokenizers = {}
+
+class Tokenizer_Http():
+    def __init__(self):
+        model_id = "qwen2.5_tokenizer"
+        self.tokenizer = AutoTokenizer.from_pretrained(model_id)
+        self.messages = [
+            {"role": "system", "content": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
+        ]
+        self.token_ids = []
+        
+        self.token_ids_cache = []
+
+    def encode(self, prompt, last_reply=None):
+        if last_reply is not None:
+            self.messages.append({"role": "assistant", "content": last_reply})
+            text = self.tokenizer.apply_chat_template(
+                self.messages,
+                tokenize=False,
+                add_generation_prompt=True
+            )
+            # print("生成的文本:\n============\n", text, "============\n")
+            self.token_ids = self.tokenizer.encode(text)[:-3]
+        self.messages.append({"role": "user", "content": prompt})
+        
+        text = self.tokenizer.apply_chat_template(
+            self.messages,
+            tokenize=False,
+            add_generation_prompt=True
+        )
+        print("生成的文本:\n============\n", text, "============\n")
+        token_ids = self.tokenizer.encode(text)
+        # 找出新增部分
+        diff = token_ids[len(self.token_ids):]
+        self.token_ids = token_ids
+        print(self.decode(diff))
+        return token_ids, diff
+
+    def decode(self, token_ids):
+        self.token_ids_cache += token_ids
+        text = self.tokenizer.decode(self.token_ids_cache)
+        if "\ufffd" in text:
+            print("text 中包含非法字符")
+            return ""
+        else:
+            self.token_ids_cache.clear()
+            return text
+        
+
+    @property
+    def bos_id(self):
+        return self.tokenizer.bos_token_id
+
+    @property
+    def eos_id(self):
+        return self.tokenizer.eos_token_id
+    
+    @property
+    def bos_token(self):
+        return self.tokenizer.bos_token
+
+    @property
+    def eos_token(self):
+        return self.tokenizer.eos_token
+    
+    def reset(self, system_prompt="You are Qwen, created by Alibaba Cloud. You are a helpful assistant."):
+        self.messages = [
+            {"role": "system", "content": system_prompt},
+        ]
+        text = self.tokenizer.apply_chat_template(
+            self.messages,
+            tokenize=False,
+            add_generation_prompt=True
+        )
+        token_ids = self.tokenizer.encode(text)[:-3]
+        self.token_ids = token_ids
+        print(self.decode(token_ids))
+        return token_ids
+        
+
+class Request(BaseHTTPRequestHandler):
+    timeout = 5
+    server_version = 'Apache'
+
+    def do_GET(self):
+        print("GET 请求路径:", self.path)
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json")
+        self.end_headers()
+
+        # 新增接口：获取 uid
+        if '/get_uid' in self.path:
+            new_uid = str(uuid.uuid4())
+            print("新 uid:", new_uid)
+            # 为该 uid 创建一个新的 Tokenizer_Http 实例
+            tokenizers[new_uid] = Tokenizer_Http()
+            msg = json.dumps({'uid': new_uid})
+        elif '/bos_id' in self.path:
+            # 获取 uid 参数（例如 ?uid=xxx）
+            uid = self.get_query_param("uid")
+            instance: Tokenizer_Http = tokenizers.get(uid)
+            if instance is None:
+                msg = json.dumps({'error': 'Invalid uid'})
+            else:
+                bos_id = instance.bos_id
+                msg = json.dumps({'bos_id': bos_id if bos_id is not None else -1})
+        elif '/eos_id' in self.path:
+            uid = self.get_query_param("uid")
+            instance: Tokenizer_Http = tokenizers.get(uid)
+            if instance is None:
+                msg = json.dumps({'error': 'Invalid uid'})
+            else:
+                eos_id = instance.eos_id
+                msg = json.dumps({'eos_id': eos_id if eos_id is not None else -1})
+        else:
+            msg = json.dumps({'error': 'Invalid GET endpoint'})
+
+        print("响应消息:", msg)
+        self.wfile.write(msg.encode())
+
+    def do_POST(self):
+        content_length = int(self.headers.get('content-length', 0))
+        data = self.rfile.read(content_length).decode()
+        print("POST 请求路径:", self.path)
+        print("接收到的数据:", data)
+        req = json.loads(data)
+
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json")
+        self.end_headers()
+
+        if '/encode' in self.path:
+            # 请求数据中必须包含 uid, text, 和可选的 last_reply
+            uid = req.get('uid')
+            prompt = req.get('text')
+            last_reply = req.get('last_reply')
+            instance: Tokenizer_Http = tokenizers.get(uid)
+            if instance is None:
+                msg = json.dumps({'error': 'Invalid uid'})
+            else:
+                token_ids, diff = instance.encode(prompt, last_reply)
+                msg = json.dumps({'token_ids': token_ids, 'diff': diff})
+        elif '/decode' in self.path:
+            uid = req.get('uid')
+            token_ids = req.get('token_ids')
+            instance: Tokenizer_Http = tokenizers.get(uid)
+            if instance is None:
+                msg = json.dumps({'error': 'Invalid uid'})
+            else:
+                text = instance.decode(token_ids)
+                msg = json.dumps({'text': text})
+        elif '/reset' in self.path:
+            uid = req.get("uid")
+            system_prompt = req.get("system_prompt")
+            instance: Tokenizer_Http = tokenizers.get(uid)
+            if instance is None:
+                msg = json.dumps({'error': 'Invalid uid'})
+            else:
+                if system_prompt is not None:
+                    print("system_prompt:", system_prompt)
+                    token_ids = instance.reset(system_prompt)
+                    msg = json.dumps({'token_ids': token_ids})
+                else:
+                    token_ids = instance.reset()
+                    msg = json.dumps({'token_ids': token_ids})
+        else:
+            msg = json.dumps({'error': 'Invalid POST endpoint'})
+
+        print("响应消息:", msg)
+        self.wfile.write(msg.encode())
+
+    def get_query_param(self, key):
+        """
+        辅助函数：从 GET 请求的 URL 中获取查询参数的值
+        例如：/bos_id?uid=xxx
+        """
+        from urllib.parse import urlparse, parse_qs
+        query = urlparse(self.path).query
+        params = parse_qs(query)
+        values = params.get(key)
+        return values[0] if values else None
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--host', type=str, default='0.0.0.0')
+    parser.add_argument('--port', type=int, default=12345)
+    args = parser.parse_args()
+
+    host = (args.host, args.port)
+    print('Server running at http://%s:%s' % host)
+    server = HTTPServer(host, Request)
+    server.serve_forever()

run_qwen2.5_0.5b_gptq_int4_ax650.sh

@@ -1,14 +1,11 @@
-./main_prefill \
+./main_ax650 \
+--system_prompt "You are Qwen, created by Alibaba Cloud. You are a helpful assistant." \
 --template_filename_axmodel "qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l%d_together.axmodel" \
 --axmodel_num 24 \
---tokenizer_type 2 \
---filename_tokenizer_model "http://127.0.0.1:12345" \
---bos 0 --eos 0 \
+--url_tokenizer_model "http://127.0.0.1:12345" \
 --filename_post_axmodel "qwen2.5-0.5b-gptq-int4-ax650/qwen2_post.axmodel" \
 --filename_tokens_embed "qwen2.5-0.5b-gptq-int4-ax650/model.embed_tokens.weight.bfloat16.bin" \
 --tokens_embed_num 151936 \
 --tokens_embed_size 896 \
---use_mmap_load_embed 1 \
---live_print 1 \
---continue 1 \
---prompt "$1"
+--use_mmap_load_embed 0 \
+--live_print 1

run_qwen2.5_0.5b_gptq_int4_axcl_aarch64.sh

@@ -1,14 +1,11 @@
 ./main_axcl_aarch64 \
 --template_filename_axmodel "qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l%d_together.axmodel" \
 --axmodel_num 24 \
---tokenizer_type 2 \
---filename_tokenizer_model "http://127.0.0.1:12345" \
---bos 0 --eos 0 \
+--url_tokenizer_model "http://127.0.0.1:12345" \
 --filename_post_axmodel "qwen2.5-0.5b-gptq-int4-ax650/qwen2_post.axmodel" \
 --filename_tokens_embed "qwen2.5-0.5b-gptq-int4-ax650/model.embed_tokens.weight.bfloat16.bin" \
 --tokens_embed_num 151936 \
 --tokens_embed_size 896 \
 --use_mmap_load_embed 0 \
 --live_print 1 \
---continue 1 \
---prompt "$1"
+--devices 0

run_qwen2.5_0.5b_gptq_int4_axcl_x86.sh

@@ -1,14 +1,11 @@
 ./main_axcl_x86 \
 --template_filename_axmodel "qwen2.5-0.5b-gptq-int4-ax650/qwen2_p128_l%d_together.axmodel" \
 --axmodel_num 24 \
---tokenizer_type 2 \
---filename_tokenizer_model "http://127.0.0.1:12345" \
---bos 0 --eos 0 \
+--url_tokenizer_model "http://127.0.0.1:12345" \
 --filename_post_axmodel "qwen2.5-0.5b-gptq-int4-ax650/qwen2_post.axmodel" \
 --filename_tokens_embed "qwen2.5-0.5b-gptq-int4-ax650/model.embed_tokens.weight.bfloat16.bin" \
 --tokens_embed_num 151936 \
 --tokens_embed_size 896 \
 --use_mmap_load_embed 0 \
 --live_print 1 \
---continue 1 \
---prompt "$1"
+--devices 0